some question about the metric accuracy?

[Zhangjyhhh]

Hi, thanks for your excellent work? I run the evaluator.py on maicity dataset . I get the similar result as your result,except accuracy. my accuracy is 3.3 that corresponds to "MAE_accuracy". I'm not sure that is "MAE_accuracy" correspond to the accuracy?

[ljjTYJR]

I guess the issue comes from the selection of ground truth. Refer to https://github.com/PRBonn/SHINE_mapping/issues/3#issuecomment-1367416983, which uses "Interaction point cloud" when evaluating accuracy

[ljjTYJR]

But I have a question that I used the provided configure file, I can not reproduce the same result as perfect as the result in https://github.com/PRBonn/SHINE_mapping/issues/3#issuecomment-1367416983

[Zhangjyhhh]

I use the "gt_map_pc_mai_inter_croped.ply" as ground truth, except "MAE_accuracy",other metric is better than the paper's metric.maybe you can use the /config/configicra2023 @ljjTYJR . so the MAE_accuracy is the accuracy ?

[ljjTYJR]

This is the reconstruction in https://github.com/PRBonn/SHINE_mapping/issues/3#issuecomment-1367416983

This is the reconstruction I use the https://github.com/PRBonn/SHINE_mapping/blob/master/config/config_icra2023/maicity_batch.yaml In fact, I think the reconstruction with the above configuration file is already good. But I wonder how can I get the first one? It is really nice!

[Zhangjyhhh]

I have no idea， maybe you can ask the author, I wish you can figure out

[YuePanEdward]

Thanks for your interest in our project and sorry for the late reply. We figured out that the key parameters for a shaper reconstruction are a smaller surface_sample_range_m and a larger free_sample_n. We will update the config file soon.